Table 1 Fault detection and localization using a hybrid deep learning model.
From: Joint processing technology of laser radar and optical image for power distribution
Input: \(\:{F}_{fused},\:{s}_{d}\:and\:GIS\:map\) |
Output: \(\:{G}_{f}\:and\:{y}^{*}\) |
Initialize deep parameters \(\:w\:and\:b\:\) |
For each layer, k in \(\:\left\{\text{1,2},\dots\:L\right\}\) |
              Compute hidden layer activation using \(\:{h}_{k}=\sigma\:\left({W}_{k}{h}_{k-1}+{b}_{k}\right)\) |
              Compute final probability using softmax function \(\:P(y\mid\:{F}_{fused})=\frac{exp({W}_{o}{h}_{L}+{b}_{o})}{{\sum\:}_{j=1}^{C}exp({W}_{o}{h}_{L}+{b}_{o})}\) |
Defined decision tree nodes using thresholding \(\:{y}^{*}=D\left(P\left({F}_{fused}\right)\right)\) |
Extract \(\:{P}_{loc}\:from\:{s}_{d}\) |
Transform \(\:{p}_{i}\) into geospatial coordinates using \(\:{G}_{f}=H\cdot\:{P}_{loc}\) |
Return \(\:{G}_{f}\:and\:{y}^{*}\) |
